Notes 10

A Very Brief Introduction to

Boundary Value Problems (BVPs)

10.1 Some Simple Boundary Value Problems for ODEs

Given a second order linear ODE, one may consider the following boundary value
problem ( ′′
y + p(x)y ′ + q(x)y = g(x)
, α≤x≤β
y(α) = y0 , y(β) = y1
The boundary value problem is called homogeneous if g(t) = 0, and y0 = y1 = 0,
and nonhomogeneous if otherwise.
As the general solution of the 2nd order linear ODE has the form

y(t) = C1 y1 (x) + C2 y2 (x) + yp (x)

the boundary value problem is a matter of finding the coefficients C1 and C2 to satisfy the
boundary conditions. However, unlike initial value problems, boundary value problems
may not have solution.
Example (Nonhomogeneous BVP) y ′′ + 4y = 0, y(0) = 2, y(π) = 0.
The general solution of y ′′ + 4y = 0 is given by y(x) = C1 cos 2x + C2 sin 2x, but no
such function can satisfy the boundary conditions:

y(0) = C1 = 2 and y(π) = C1 cos 2π + C2 sin 2π = C1 = 0

i.e., C1 = 2, and C1 = 0, which is not possible.

Example (Nonhomogeneous BVP) y ′′ + 4y = 0, y(0) = 2, y(π) = a.
Similarly, y(x) = C1 cos 2x + C2 sin 2x, with boundary conditions:

y(0) = C1 = 2, and y(π) = C1 cos 2π + C2 sin 2π = C1 = a

i.e., there is no solution if a 6= 2.

However, if a = 2, there are infinitely many solutions:

y(t) = 2 cos 2x + C2 sin 2x

71
 72

where C2 is an arbitrary constant.

Example (Nonhomogeneous BVP) y ′′ + 4y = 0, y(0) = 2, y( π4 ) = a.
y(t) = C1 cos 2x + C2 sin 2x, with boundary conditions:
π π π
y(0) = C1 = 2, y( ) = C1 cos + C2 sin = C2 = a
4 2 2
i.e., C1 = 2 and C2 = a, and there is a unique solution for the boundary value problem:

y(t) = 2 cos 2x + a sin 2x

The examples examples above show that in general, a BVP may have: (i) no solution,
(ii) a unique solution, or (iii) infinitely solutions.
Question: Can there be 3, or 4, or a finite number of solutions for such BVPs?

The answer is no. If φ(x) and ψ(x) are two solutions of the BVP
( ′′
y + p(x)y ′ + q(x)y = g(x)
, α≤x≤β
y(α) = y0 , y(β) = y1

then their difference φ − ψ must be a solution of the homogeneous boundary value

problem  2
 d y2 + p(x) dy + q(x)y = 0
dx dx
 y(α) = 0, y(β) = 0
which has at least one obvious solution, namely, the trivial solution y(x) ≡ 0.
Hence if the trivial solution is the only possible solution for the corresponding homo-
geneous BVP, then φ − ψ ≡ 0, i.e., the original BVP has a unique solution.
On the other hand, if the homogeneous BVP has a non-trivial solution y(x) 6=≡ 0,
then for any constant c, cy(x) is still a solution of the homogeneous BVP; i.e., there are
infinitely many solutions for the homogeneous BVP in this case.

Consequently, a nonhomogeneous BVP

( ′′
y + p(x)y ′ + q(x)y = g(x)
, α≤x≤β
y(α) = y0 , y(β) = y1

may have
(i) no solution;
(ii) a unique solution φ(t), if the corresponding homogeneous BVP has only the
trivial solution;
(iii) infinitely many solutions; if the corresponding homogeneous BVP has non-
trivial solutions.
In the last case, the general solution of the nonhomogeneous BVP has the form y(x) =
φ(x) + h(x), where φ(x) is one particular solution of the nonhomogeneous BVP, and h(x)
is the general solution of the corresponding homogeneous BVP.
 73

10.2 Eigenvalues and Eigenfunctions of BVPs

We consider here the simplest homogeneous BVP problems
 2
 d y
+ λy = 0
dx2 (*)
 y(α) = 0 = y(β)

d2 y
If we let L[y] = − 2 , the equation can be expressed as L[y] = λy.
dx
In analogy with eigenvalues and eigenvectors of matrices, we called λ an eigenvalue
of the boundary value problem (*) if it has a non-trivial solution y(x) 6≡ 0.
The corresponding non-trivial solutions, will then be called the eigenfunctions of
the BVP.
Example The homogeneous BVP y ′′ + 3y = 0, y(0) = 0, y(π) = 0, has only the trivial
solution, and hence 3 is not an eigenvalue of the BVP (*).
√ √
In fact, if y(x) = C1 cos 3 x + C2 sin 3 x is a solution of the differential equation
with boundary conditions
√ √
y(0) = C1 = 0 and y(π) = C1 cos 3π + C2 sin 3π = 0

then C1 = C2 = 0, and therefore the trivial solution y(x) = 0 is the only solution of the
homogeneous BVP.
Example The homogeneous BVP y ′′ + 4y = 0, y(0) = 0, y(π) = 0 has non-trivial
solutions.
Consider y(x) = C1 cos 2t + C2 sin 2t, with boundary conditions

y(0) = C1 = 0 and y(π) = C1 cos 2π + C2 sin 2π = C1 = 0

i.e., C1 = 0, and C2 can be any constant. Here the BVP has infinitely many solutions

y(x) = C2 sin 2x

In the language of eigenvalue-eigenfunction, 4 is an eigenvalue of the BVP (*), and

C sin 2x, with C 6= 0, are the corresponding eigenfunctions.

Recall that the general solution of the equation y ′′ + λy = 0 is:



 C1 + C2 x if λ = 0
 √ √
y(x) = C1 e− −λ x + C2 e −λ x if λ < 0

 √ √

C1 cos λ x + C2 sin λ x if λ > 0

Given the boundary conditions y(0) = y(L) = 0, we have:

• Case (i) λ = 0.
Then y = C1 + C2 x can not vanish at two points y(0) = 0 and y(L) = 0, unless
C1 = C2 = 0. Only trivial solution: y(x) = 0.
 74

• Case (ii) λ < 0.

√ √
Then y(x) = C1 e− −λ x + C2 e −λ x can not vanish at two points, since y(0) =
y(L) = 0 implies
(
C1 + C2 = 0
√ √ ⇐⇒ C1 = C2 = 0
C1 e− −λ L + C2 e −λ L =0

Hence the BVP has no non-trivial solution when λ < 0.

• Case (iii) λ > 0.

√
If λ > 0, y(0) = y(L) = 0 implies C1 = 0 and C2 sin λ L = 0.
√
In particular, non-trivial solutions exist when C2 =
6 0 but sin λL = 0, i.e., when
√
λL = nπ, where n = 0, ±1, ±2, . . .

Therefore the eigenvalues of the BVP (*) are

n2 π 2
λ= , where n = 0, 1, 2, . . .
L2
n2 π 2
The eigenfunctions corresponding to λ = L2
are
nπx
y(x) = C sin
L
where C 6= 0 is an arbitrary constant.

Exercise By similarly considerations, it is not hard to find the following eigenvalues

and eigenfunctions for some other kinds of boundary value problems:

BVP’s for the equation y ′′ + λy = 0

For what λ can the following BVPs have non-trivial solutions ?

boundary y(0) = 0 y ′ (0) = 0 y(−L) = y(L)

conditions y(L) = 0 y ′ (L) = 0 y ′ (−L) = y ′ (L)
n2 π 2 n2 π 2 n2 π 2
eigenvalues λn L2 L2 L2
n = 1, 2, 3, . . . n = 0, 1, 2, 3, . . . n = 0, 1, 2, 3, . . .
basic sin nπx
sin nπx cos nπx L
eigenfunctions L L cos nπx
L

10.3 Solving Certain Nonhomogeneous BVP by Eigenfunc-

tions
Recall that
( BVP eigenvalues: λn = n2 ,
y ′′ + λy = 0 =⇒ eigenfunctions: sin nx
y(0) = y(π) = 0 n = 1, 2, 3, . . .
 75

Let’s see how one could solve some nonhomogeneous BVP by the superposition of
eigenfunctions.
Example Solve the following nonhomogeneous BVP:

y ′′ + 3y = 3 sin x − 2 sin 2x + 7 sin 3x, y(0) = y(π) = 0

Note that the function on the right hand side is a linear combination of the eigenfunctions
sin x, sin 2x, sin 7x of the BVP

L[y] + λy = 0, y(0) = 0 = y(π) (**)

corresponding to the eigenvalues 12 , 22 , 32 respectively.

What if we try a suitable linear combination of the these eigenfunctions?
Let
y(x) = b1 sin x + b2 sin 2x + b3 sin 3x
Obvious such a y(x) would satisfy the boundary conditions. Putting that into the equa-
tion, we have

(−b1 + 3b1 ) sin x + (−4b2 + 3b2 ) sin 2x + (−9b3 + 3b3 ) sin 3x = 3 sin x − 2 sin 2x + 7 sin 3x

Hence by comparing coefficients, we have

 
3
 2b1 = 3
  b1 = 2

−b2 = −2 ⇐⇒ b2 = 2

 −6b = 7 
 b = −7
3 3 6

Thus a solution of the non-homogeneous BVP is

3 7
y(t) = sin x + 2 sin 2x − sin 3x
2 6
There is no other solution, since the homogeneous BVP y ′′ + 3y = 0, y(0) = y(π) = 0
has only trivial solution, as 3 is not an eigenvalue of the BVP.
This method in the example above can be easily generalized to the nonhomogeneous
BVP
y ′′ + λy = f (x), y(0) = y(π) = 0
where λ is not an eigenvalue of the corresponding homogeneous BVP

y ′′ + λy = 0, y(0) = y(π) = 0
n
X
and f (x) = bn sin nx is a linear combination of the eigenfunctions of the corresponding
n=1
homogeneous BVP.
More generally, we may even consider the case of

f (x) = superposition of infinitely many eigenfunctions

 76

by the same trick. More precisely, for any λ 6= 12 , 22 , 32 , . . . , k2 , . . ., the BVP

∞
X
′′
y + λy = bk sin kx, y(0) = y(π) = 0
k=1

can be solved by putting

y(x) = c1 sin x + c2 sin 2x + c3 sin 3x + · · ·

into the differential equation, which will lead to

bk
−k2 ck + λck = bk ⇔ ck =
−k2 +λ
Of course, one would need to work with functions which can be expressed as an infinite
sine/consine series.

10.4 A Brief Description of Fourier Series

There are some nice relations between those eigenfunctions, sin nπx nπx
L , cos L , arising form
various BVPs, namely, that they are mutually orthogonal, in the sense that
Z L (
mπx nπx 0 if m 6= n
cos cos dx =
−L L L L if m = n

Z (
L
mπx nπx 0 if m 6= n
cos sin dx =
−L L L L if m = n
Z (
L
mπx nπx 0 if m 6= n
sin sin dx =
−L L L L if m = n
Now, if f (x) is a sine/cosine series (with convergence assumed),
∞
a0 X mπx mπx

f (x) = + am cos + bm sin
2 L L
m=1

using the orthogonality conditions above, we can multiply cos nπx nπx
L , or sin L , to both
sides and then integrate to find the coefficients
Z L
1 nπx
an = f (x) cos dx, n = 0, 1, 2, . . .
L −L L
Z L
1 nπx
bn = f (x) sin dx, n = 1, 2, 3, . . .
L −L L
assuming that term by term integration is okay.
In general, the trigonometric series
∞
a0 X mπx mπx

+ am cos + bm sin
2 L L
m=1
 77

thus obtained from a given function f (x) defined on −L ≤ x ≤ L is called the Fourier
series of f (x).
Generally speaking, after writing down the Fourier series of a function f (x), there is
a problem of convergence: is the series going to converge back to the function f (x) on the
interval −L ≤ x ≤ L ?

Theorem Suppose that f and f ′ are piecewise continuous on the interval −L ≤ x ≤ L,

and extend f to the whole real line periodically, with period 2L. The the Fourier series
of f (x) converges to f (x) at all points where f is continuous, i.e.,
∞
a0 X mπx mπx

f (x) = + am cos + bm sin
2 m=1
L L

when x is a point of continuity, and to f (x+)+f

2
(x−)
at all points where f is discontinuous,
i.e.
∞
f (x+) + f (x−) a0 X mπx mπx

= + am cos + bm sin
2 2 L L
m=1

when there is a jump in function value f (x+) − f (x−) 6= 0.

Exercise Show that the Fourier series of
(
0, −L < x < 0
f (x) =
L, 0 < x < L

is 
∞
L 2L X sin
(2n−1)πx  f (x) if x 6= 0, ±nL
L
+ =
2 π 2n − 1  L if x = 0, ±nL
n=1
2
A function f (x) defined on the real line is called an even function, if f (−x) = f (x)
for any x. It is called an odd function if f (−x) = −f (x). It is a straightforward
calculation that
Z L
nπx nπx
f (x) even =⇒ f (x) sin odd =⇒ f (x) sin dx = 0
L −L L

and Z L
nπx nπx
f (x) odd =⇒ f (x) cos odd =⇒ f (x) cos dx = 0
L −L L
Thus we have

(i) If f (x) is even, periodic with period 2L, then its Fourier series is a Fourier cosine
series:
∞
a0 X nπx
f (x) ∼ + an cos
2 n=1
L
where Z L
2 nπx
an = f (x) cos dx
L 0 L
 78

(ii) If f (x) is odd, periodic with period 2L, then its Fourier series is a Fourier sine
series:
∞
X nπx
f (x) ∼ bn sin
n=1
L
where Z L
2 nπx
bn = f (x) sin dx
L 0 L

Starting from a function f (x) on 0 < x < L, we can get either a Fourier Cosine, or
Sine series by:

(i) Extend f (x) as an even function of period 2L, then

∞
a0 X nπx
f (x) ∼ + an cos
2 L
n=1

(ii) Extend f (x) as an odd function of period 2L, then

∞
X nπx
f (x) ∼ bn sin
n=1
L

Example f (x) = x, where 0 < x < π.

(i) Extend f (x) as an even function of period 2π, which is always continuous:
(
x 0<x<π
f (x) =
−x −π < x < 0

Then Z (
π π n=0
2
an = x cos nxdx = 2((−1)n −1)
π 0 n2 π
n = 1, 2, . . .
∞
π 4X 1
f (x) = − cos(2n − 1)x
2 π (2n − 1)2
n=1

Taking x = 0, which is a point of continuity, we have

π 4 1 1 1

0= − 2
+ 2 + 2 + ···
2 π 1 3 5
∞
X 1 1 1 1 π2
2
= 2 + 2 + 2 + ··· =
n=1
(2n − 1) 1 3 5 8

Noting that
1 1 1 1 1 1 1 1  1 1 1 
+ + + + · · · = 1 + + + + · · · + + + + · · ·
12 22 32 42 4 22 32 42 12 32 52
we have  π2
1 1 1 4 1 1 1
1+ + + + · · · = + + + · · · =
22 32 42 3 12 32 52 6
 79

(ii) Extending f (x) as an odd function of period 2π, we have f (x) = x. Hence
Z
2 π 2 cos nπ 2(−1)n+1
bn = x sin nxdx = − =−
π 0 n n
∞
X 2(−1)n+1
f (x) ∼ sin nx
n
n=1

Taking x = π2 , which is a point of continuity, we have

π 1 1 1

= 2 1 − + − + ···
2 3 5 7
∞
X (−1)n+1 1 1 1 π
=1− + − + ··· =
n=1
2n − 1 3 5 7 4
At a point of discontinuity, say x = π, we have
∞
f (π+) + f (π−) −1 + 1 X 2(−1)n+1
= =0= sin n(0)
2 2 n=1
n

Example Solve the boundary value problem

y ′′ + 2y = x, y(0) = 0 = y(π)

Extending the function x as an odd function of period 2π, we have

 1 1 1 
y ′′ + 2y = 2 sin x − sin 2x + sin 3x + sin 4x + · · ·
2 3 4
Consider a sine series expansion of y(x) as

y(x) = b1 sin x + b2 sin 2x + b3 sin 3x + · · ·

and put it into the equation, we have

 1 1 1 
(−b1 +2b1 ) sin x+(−4b2 +2b2 ) sin 2x+(−9b3 +2b3 ) sin 3x+· · · = 2 sin x− sin 2x+ sin 3x+ sin 4x+· · ·
2 3 4
1 2(−1)n+1 2(−1)n+1
bn = · =
−n2 + 2 n n(2 − n2 )
∞ ∞ 
X 2(−1)n+1 X (−1)n+1 n(−1)n+1 
y(x) = sin nx = + sin nx
n(2 − n2 ) n 2 − n2
n=1 n=1
∞
x X n(−1)n+1
y(x) = + 2
sin nx
2 n=1
2 − n
| {z }
√
Fourier series of −π√ sin( 2x)
2 sin( 2π)

In fact, the general solution of the equation y ′′ + 2y = x can easily be found as

√ √
y(t) = x2 + C1 cos 2x + C2 sin 2x. Putting in the boundary conditions, you could fix
C1 = 0, C2 = − 2 sin(π√2π) ; same as the one found by the method of Fourier expansion.
 80

10.5 A Brief Introduction to the Method of Separation of

Variables of Some Basic Linear Partial Differential
Equations
Consider the temperature u(x, t) at the point x (or more precisely the cross section
at x) on a thin uniform rod of length L. Assuming there is no heat exchange with the
surroundings, and both end of the rod are kept at temperature 0, then the heat conduction
in the rod can be modeled by the heat equation as follows:

∂2u ∂u
α2 = , 0 < x < L, t > 0
∂x2 ∂t
u(0, t) = 0, u(L, t) = 0, ,
u(x, 0) = f (x),

where f (x) is the initial temperature distribution in the rod. The constant α2 is called
the thermal diffusivity.
The idea of the method of separation of variables is to consider u(x, t) = X(x)T (t)
as a product of a function of x and a function of t. Putting u(x, t) into the heat equation,
we have
X ′′ 1 T′
α2 X ′′ (x)T (t) = X(x)T ′ (t) ⇐⇒ = 2·
X α T
Since the left hand side is a function of x, and the right hand side is a function of t,
both must be actually a constant function. Hence there is a constant −λ so that
(
X ′′ 1 T′ X ′′ + λX = 0
= 2· = −λ ⇐⇒
X α T T ′ + α2 λT = 0

When u(x, t) is not the zero function, the boundary conditions u(0, t) = X(0)T (t) = 0
and u(L, t) = X(L)T (t) = 0 imply the boundary conditions X(0) = 0, X(L) = 0.

′′
 X + λX = 0

2
X(0) = X(L) = 0 and T = Ce−α λt

 (from boundary condition)

Now, the boundary value problem for X(x) has non-trivial solutions X(x) = B sin nπx L
2 2
if and only if λ = nLπ2 , n = 1, 2, 3, . . ..
2 2 2 2
Hence for each n = 1, 2, 3, . . ., we have the basic solution un (x, t) = e−n π α t/L sin nπx
L
of the heat equation
∂2u ∂u
α2 2 =
∂x ∂t
Moreover, their infinite linear combinations
∞
X
u(x, t) = cn un (x, t)
n=1

will also satisfy the heat equation and the boundary condition u(0, t) = 0 = u(L, t).
 81

Now we only need to choose suitable coefficients cn ’s so that u(x, t) will satisfy also
the initial condition
∞
X nπx
u(x, 0) = cn sin = f (x)
n=1
L

By expanding f (x) to its Fourier sine series, we have

Z L
2 nπx
cn = f (x) sin dx
L 0 L

Example Solve the heat equation for 0 < x < 30, t > 0, under the given conditions:

uxx = ut , u(0, t) = 0 = u(30, t), u(x, 0) = 40 − 3x

∞
X 2 π 2 t/900 nπx
Just consider u(x, t) = cn e−n sin and compare at t = 0 with the
n=1
30
Fourier sine series of f (x) = 40 − 3x so that
Z 30
2 nπx
cn = (40 − 3x) sin dx
30 0 30
Z 30
1 30 nπx
= (3x − 40)d cos dx
15 0 nπ 30
Z 30
2 h nπx i30 6 nπx
= (3x − 40) cos − cos dx
nπ 30 0 nπ 0 30
Hence
20[5(−1)n + 4]
cn =
nπ
and
∞
X 20[5(−1)n + 4] 2 π 2 t/900 nπx
u(x, t) = e−n sin
nπ 30
n=1

Example Solve the heat equation for 0 < x < 30, t > 0, under the given nonhomoge-
neous boundary conditions:

uxx = ut , u(0, t) = 20, u(30, t) = 50, u(x, 0) = 60 − 2x

The trick is to consider a solution of the steady-state temperature u(x, t) = v(x) (i.e.,
independent of time) which satisfies the non-homogeneous boundary conditions:

v ′′ (x) = 0, v(0) = 20, v(30) = 50

The solution can be found easily as v(x) = x + 20.

Now, consider w(x, t) = u(x, t) − v(x), which reduces the original equation for u(x, t)
into

wxx = wt , w(0, t) = 0, w(L, t) = 0, w(x, 0) = u(x, 0) − v(x) = 40 − 3x

The equation for w has been solved in the previous example.

 82

Example - Heat Conduction in a Bar with Insulated Ends

Solve the heat equation for 0 < x < 25, t > 0, under the given boundary conditions:

uxx = ut , ux (0, t) = 0, ux (25, t) = 0, u(x, 0) = x

Here the boundary conditions ux (0, t) = 0 and ux (25, t) = 0 mean that there is no
heat flow at both ends of the bar.

• Consider u(x, t) = X(x)T (t) again.

• Since the boundary conditions are different, the boundary value problem for X(x)
is: (
X ′′ + λX = 0
, and T (t) = Ce−λt
X ′ (0) = 0 = X ′ (25)

n2 π 2
The eigenvalues are now λn = 252
, n = 0, 1, 2, . . ., with eigenfunctions cos nπx
25 .

• By the same idea of superposition, pick suitable coefficients cn such that

∞
c0 X 2 2 nπx
u(x, t) = + cn e−n π t/625 cos
2 25
n=1

At t = 0, we consider
∞
c0 X nπx
u(x, 0) = + cn cos = Fourier cosine series of x
2 n=1
25

i.e., Z 25
2 nπx
cn = x cos dx
25 0 25
• The solution is
∞
25 X 50[(−1)n − 1] −n2 π2 t/625 nπx
u(x, t) = + 2 2
e cos
2 n π 25
n=1

10.6 Wave and Laplace Equations via Separation of Vari-

ables
Similarly, the same trick of separation of variables can be used to solve some other equa-
tions with appropriate boundary/initial conditions:
• Wave Equations
For example, the small vibration of a string of length L fixed at both end can be
modeled by:

2
 α uxx = utt

u(0, t) = 0, u(L, t) = 0 (t ≥ 0)

 u(x, 0) = f (x), u (x, 0) = g(x)
t (0 ≤ x ≤ L)

Let u(x, t) = X(x)T (t) again, and see what sorts of boundary value problems one
has to solve.
 83

• Laplace Equations
For example, the steady-state heat conduction of a rectangular plate can be modeled
by: 
 uxx + uyy = 0
 (0 < x < a, 0 < y < b)
u(x, 0) = 0, u(x, b) = 0 (0 < x < a)

 u(0, y) = 0, u(a, y) = f (y) (0 < y < b)

Let u(x, y) = X(x)Y (y), and see what sorts of boundary value problems one has.

Roughly speaking, by letting u(x, t) = X(x)T (t) or u(x, y) = X(x)Y (y) respectively
for the two equations above, one would end up with different boundary value problems for
finding the ‘eigenfunctions’ Xn , Tn or Yn . Then the solution can be found by superposition
and Fourier expansion:
X X
cn Xn (x)Tn (t) or cn Xn (x)Yn (y)

Details of heat equations and Laplace equations are treated in courses on partial
differential equations, such as MATH4052.

10.7 More General Two-Point Boundary Value Problems

In previous sections, we considered the eigenvalues/eigenfunctions of the operator L =
d2
− 2 with boundary condition y(α) = 0 = y(β). In fact, we can consider more general
dx
boundary conditions, as well as more general differential operators.

d2
Example L=− with more general boundary conditions.
dx2

 d2 y

 L[y] = λy ⇐⇒ + λy = 0

 dx2
ay(0) + by ′ (0) = 0 , 0≤x≤L

 cy(L) + dy ′ (L) = 0



(assuming not all a, b, c, d are zero)

Here the boundary conditions mix the boundary values of y and y ′ together.
These are considered as ‘easy’ boundary value problems, simply because we know
exactly the general solution of the equation y ′′ + λy = 0, and hence can check easily
whether there are non-trivial solutions of the BVP.
Example Find all eigenvalues and eigenfunctions of the BVP

y ′′ + λy = 0, y(0) = 1, y(π) + y ′ (π) = 0

If λ = 0, then y = c1 + c2 x. y(0) = 0 =⇒ c1 = 0. Then

0 = y(π) + y ′ (π) = c2 π + c2 =⇒ c2 = 0
 84

i.e, no non-trivial solution,

√
and λ = √ 0 is not an eigenvalue of the BVP.
If λ < 0, then y = c1 e −λx +c2 e −λx . Check that there is also no non-trivial solution
−

in this case. (Exercise)

√ √
If λ > 0, then y = c1 cos λx + c2 sin λx.

y(0) = 0 =⇒ c1 = 0
√ √ √
Hence y = c2 sin λx and y ′ = λc2 cos λx. The condition y(π) + y ′ (π) = 0 is then
√ √ √
c2 sin λπ + λc2 cos λπ = 0
√ √
λ = − tan λπ
By looking at the graph of the function f (x) = x and g(x) = − tan πx, we have the
rough estimate
(2n − 1)π p
< λn < nπ
2
and
(2n − 1)2 π 2
λn ≈ for large n
4

“Harder BVP Problems”

Many BVPs in applications are related to more general linear differential operators
of 2nd order:  h 
1 d di
L=− p(x) − q(x)
r(x) dx dx
where p(x), p′ (x), q(x), r(x) are all continuous on the interval 0 ≤ x ≤ L, with p(x) > 0,
r(x) > 0. 
′ ′
 L[y] = λy ⇐⇒ [p(x)y ] − q(x)y + λr(x)y = 0

ay(0) + by ′ (0) = 0, cy(L) + dy ′ (L) = 0

 (assuming not all a, b, c, d are zero)

These are called the (regular) Sturm-Liouville Boundary Value Problems, which
quite often occur when solving PDEs with the method of separation of variables. They
are ‘hard’ in the sense that there is no general way to find the general solution of the
equation [p(x)y ′ ]′ − q(x)y + λr(x)y = 0.
However, the structure of the eigenvalues and eigenfunctions of such ’harder BVPs’
d2
is the same as the easier BVPs associated to L = − dx 2.

1. All the eigenvalues are real numbers.

2. The eigenvalues form an unbounded infinite increasing sequence

λ1 < λ2 < · · · < λn < · · · −→ +∞

3. For each eigenvalue λn , there is an eigenfunction φn (x) whose scalar multiples gen-
erates all eigenfunctions with λn as eigenvalue. Moreover, φn has exactly n − 1 zero
in 0 < x < L.
 85

4. Eigenfunctions with different eigenvalues are orthogonal relative to the weight

function r(x), i.e., for λn 6= λm ,
Z L
φn (x)φm (x)r(x) dx = 0
0

5. Every piecewise smooth function f (x) can be represented by a ‘generalized Fourier

series’
∞
X
f (x) ∼ an φn (x)
n=1

where the infinite series converges to [f (x+) − f (x−)]/2 for every x in the interval
0 < x < L.

Nonhomogeneous BVPs
Eigenfunction expansion (generalized Fourier series) can be used to solve nonhomo-
geneous BVPs,
(
L[y] = µy + f (x)/r(x) ⇐⇒ −[p(x)y ′ ]′ + q(x)y = µr(x)y + f (x)
ay(0) + by ′ (0) = 0, cy(L) + dy ′ (L) = 0
 h 
1 d di
where µ is not an eigenvalue of L = − p(x) − q(x) .
r(x) dx dx
Similar to the Fourier series method, we can consider superposition of eigenfunctions
P P
y = an φn , and the eigenfunction expansion of f (x)/r(x) = cn φn . It is then straight-
forward to find
cn
an =
λn − µ
i.e., the solution of such a nonhomogeneous BVP can be expressed in terms of the eigen-
functions:
∞
X cn
y= φn (x)
λ −µ
n=1 n

Boundary value problems of PDEs, or Sturm-Liouville differential operators, belong

to more advanced courses in differential equations though.